Years ago, when I worked with Avaya, it was normal to build a contact centre from products supplied by several vendors. Avaya handled the core telephony and routing. Other suppliers provided recording, workforce management and reporting. The products made sense on their own. Getting them to work together, then keeping them working through upgrades, was the hard part.
I saw firsthand how every integration added another dependency and another place for a failure to hide. When something broke, ownership was not always obvious. When all-in-one cloud suites arrived, simplification became a powerful part of the pitch: fewer moving parts, one administration layer and clearer accountability.
That history is why today's wave of standalone AI announcements feels familiar. I am not suggesting the technology is the same, but I do wonder whether we are starting to rebuild some of the fragmentation we worked so hard to remove.
Three recent moves brought the question into focus:
Genesys announced Standalone AI Agents, designed to let organizations add its Agentic Virtual Agent to an existing contact centre without first moving the full environment to Genesys Cloud.
Five9 entered into a definitive agreement to acquire Level AI in a transaction valued at approximately US$325 million. Five9 says Level AI applications will also be offered as standalone solutions for organizations using on-premises or other cloud platforms.
Microsoft put Contact Center Intelligence into public preview. It adds AI agents, quality and analytics to existing contact centres, with connectors for platforms including Five9, Cisco Webex Contact Center and Avaya.
Avaya is the name that caught my attention. An environment like the ones I worked on years ago can now become the foundation for a new AI layer instead of being replaced first.
There is a real benefit here. An organization may be able to add useful AI without committing to a multi-year contact centre migration. That can be a practical choice when the core platform is stable, the replacement budget is not available, or the business wants to learn before making a larger decision.
Modern REST APIs and cloud architecture make these connections easier than the proprietary CTI links I worked with in the Avaya era. Easier integration, however, does not guarantee simpler operations. Interfaces still change. Customer context still has to move between systems. When something fails, suppliers can still disagree about where the problem began. An API is not an operating model.
The strategy also makes commercial sense for vendors. Many buyers want to invest in AI now, even if they are not ready to replace the rest of their contact centre. Selling the AI layer separately meets that demand and keeps the door open to a broader platform sale later.
It also changes the buying decision. If an enterprise can add capable AI to an existing Cisco or Avaya estate, a full platform migration may feel less urgent. That does not make the wider suite unnecessary. It does mean vendors need to explain clearly what the rest of the platform adds.
Before adding another AI layer, I would ask five practical questions:
Who owns the context? When self-service hands a customer to a person, does the full history move with them?
Can operations see the whole journey? Can they follow one interaction across vendor boundaries?
Who governs the combined system? How are security, privacy and compliance applied across providers?
Who finds the fault? When automation fails, who traces the problem across AI, APIs and telephony?
Who owns the outcome? Is anyone accountable for the customer's result, or only for each component's uptime?
Modular AI may be a sensible next step. It gives organizations more choice and lets them modernize around systems that still work. It will feel like progress only if interoperability comes with clear ownership, end-to-end visibility and a support model that does not send the customer from vendor to vendor.